Jääskeläinen et al. (2025) High-resolution soil moisture mapping in northern boreal forests using SMAP data and downscaling techniques

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-11-13
• Authors: Emmihenna Jääskeläinen, Miska Luoto, Pauli Putkiranta, Mika Aurela, Tarmo Virtanen
• DOI: 10.5194/hess-29-6237-2025

Research Groups

• Arctic Space Centre, Finnish Meteorological Institute (FMI), Helsinki, Finland.
• Department of Geosciences and Geography, University of Helsinki, Finland.
• Ecosystems and Environment Research Programme, Faculty of Biological and Environmental Sciences, University of Helsinki, Finland.

Short Summary

This study develops a machine-learning-based downscaling model to estimate soil moisture at 1 km and 250 m spatial resolutions across northern boreal forests. By integrating SMAP satellite data with vegetation and weather parameters, the model improves soil moisture prediction accuracy over forested sites compared to original coarse-resolution products.

Objective

• To develop and validate a high-resolution (1 km and 250 m) soil moisture estimation model specifically optimized for boreal forest ecosystems using machine learning and satellite data.

Study Configuration

• Spatial Scale: Northern Finland (65.5–70.0° N) for model training and testing; boreal forest zones in Alaska for independent validation.
• Temporal Scale: Annual snow-free periods from 1 May to 15 October, spanning the years 2019–2023.

Methodology and Data

• Models used: Light Gradient-Boosting Machine (LightGBM), a leaf-wise growth-based gradient boosting framework.
• Data sources:
  • Satellite: SMAP L-band radiometer (SPL3SMP V009, 36 km resolution), MODIS (NDVI and EVI at 250 m resolution), and Global Precipitation Measurement (GPM) IMERG Final Run.
  • Reanalysis/Interpolated: FMI interpolated daily weather observations (1 km resolution), ERA5-Land 2 m air temperature.
  • Land Cover: CORINE Land Cover 2018 (100 m resolution).
  • In situ: FMI-ARC (Finland), International Soil Moisture Network (ISMN), and National Ecological Observatory Network (NEON) stations at depths of −5 cm to −6 cm.

Main Results

• The Gradient Boosting (GB) model successfully downscaled SMAP data to 1 km and 250 m while retaining large-scale spatial and temporal variability.
• Accuracy Improvement: In validation against independent Alaska forest sites, the model reduced the Root Mean Square Error (RMSE) from 0.103 m³ m⁻³ (original SMAP) to 0.092 m³ m⁻³.
• Correlation: The correlation coefficient (R) increased from 0.46 to 0.55 over forested validation sites.
• Predictor Importance: Vegetation indices (NDVI, EVI) and cumulative air temperature were the most significant predictors, while precipitation had a comparatively smaller effect.
• Uncertainty: The total model uncertainty was calculated at 0.080 m³ m⁻³.

Contributions

• Boreal Optimization: Unlike many global products, this model is specifically tuned for boreal forest environments, addressing the challenges of dense canopy interference in microwave retrievals.
• High Spatial Resolution: Achieved 250 m resolution mapping, which is significantly finer than the standard 36 km or 9 km SMAP products, facilitating better local-scale hydrological and carbon cycle studies.
• Methodological Validation: Demonstrated that combining L-band passive microwave data with optical vegetation indices and weather data via machine learning effectively mitigates the coarse resolution of radiometer data.

Funding

• Ministry of Transport and Communications through the Integrated Carbon Observing System (ICOS) Finland.
• Research Council of Finland, Luonnontieteiden ja Tekniikan Tutkimuksen Toimikunta (grant no. 347662).

Citation

@article{Jääskeläinen2025Highresolution,
  author = {Jääskeläinen, Emmihenna and Luoto, Miska and Putkiranta, Pauli and Aurela, Mika and Virtanen, Tarmo},
  title = {High-resolution soil moisture mapping in northern boreal forests using SMAP data and downscaling techniques},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6237-2025},
  url = {https://doi.org/10.5194/hess-29-6237-2025}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)